CODEWITHSUNDEEP
pythonkeras
giulia
giulia

Reputation: 111

AttributeError: module 'keras.utils' has no attribute 'get_file' using segmentation_models

I'm trying to use segmentation models but I can't fix this error. I've searched for this particular one but couldn't find an answer. I'm using pycharm and this error is linked to this specific line of code BACKBONE = 'resnet34'

model1 = sm.Unet(BACKBONE, weights=None,
                 encoder_weights='imagenet',
                 classes=num_classes,
                 activation='softmax',
                 decoder_block_type = 'upsampling')

which is also the 83rd. I searched in the documentation and apparently the versions of tensorflow keras etc satisfy the requirements.I really don't know what to do given the fact that I really tried to install and uninstall everything in many combinations in order to get this piece of code to work.Thank you all for your help and time! Below there's the complete error, hoping it might help you!

    `Traceback (most recent call last):
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\main.py", line 83, in <module>
        model1 = sm.Unet('resnet34', weights=None,
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\segmentation_models\__init__.py", line 34, in wrapper
        return func(*args, **kwargs)
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\segmentation_models\models\unet.py", line 221, in Unet
        backbone = Backbones.get_backbone(
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\segmentation_models\backbones\backbones_factory.py", line 103, in get_backbone
        model = model_fn(*args, **kwargs)
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\classification_models\models_factory.py", line 78, in wrapper
        return func(*args, **new_kwargs)
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\classification_models\models\resnet.py", line 314, in ResNet34
        return ResNet(
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\classification_models\models\resnet.py", line 280, in ResNet
        load_model_weights(model, model_params.model_name,
      File "C:\Users\Giulia\PycharmProjects\multiclass_new\venv\lib\site- 
   packages\classification_models\weights.py", line 25, in load_model_weights
        weights_path = keras_utils.get_file(
    AttributeError: module 'keras.utils' has no attribute 'get_file'

Upvotes: 11

Views: 17786

Answers (5)

Bala Sai Sigireddy
Bala Sai Sigireddy

Reputation: 11

If you are using Google Colab, I recommend running:

import os
import time
os.environ['SM_FRAMEWORK'] = 'tf.keras'
time.sleep(3)
import semantic_model as sm

or:

os.environ['SM_FRAMEWORK'] # and import statements on seperate cells

it may be there is someprocess with setting of environ variable and working i dont know how it works but it works

Upvotes: 0

Omar Essam El-Din
Omar Essam El-Din

Reputation: 1873

I had same problem but with vgg u-net model , This worked for me

!apt-get install -y libsm6 libxext6 libxrender-dev

!pip install opencv-python

!pip install git+https://github.com/divamgupta/image-segmentation-keras

from keras_segmentation.models.unet import vgg_unet

or check here how it's implemented https://gitee.com/sanyanjie/image-segmentation-keras

Upvotes: 0

Priyam Jha
Priyam Jha

Reputation: 1

To solve this issue, try importing the module EfficientNetB0 directly, as the code below:

import efficientnet.tfkeras as efn

Upvotes: 0

Hagbard
Hagbard

Reputation: 3700

The other answer provided here didn't work for me. Instead, upgrading keras did the trick for me via:

pip install --upgrade keras

Upvotes: -1

Harshil Gandhi
Harshil Gandhi

Reputation: 230

You can try:

import segmentation_models as sm

sm.set_framework('tf.keras')

sm.framework()

Worked for me on google colab!

Upvotes: 23

Related Questions